The Agent Skills Directory

[SAFE]: The skill is instructional in nature and provides guidance for the agent to optimize its internal reasoning parameters. No prompt injection attempts, safety filter bypasses, or malicious instructions were detected.
[SAFE]: No external network dependencies, remote code downloads, or unauthorized data exfiltration patterns were identified. The skill operates entirely within the agent's local environment.
[SAFE]: Data access is limited to the skill's own telemetry and command history logs (stored in ~/.claude/.metrics/reasoning/). This access is used solely for generating performance and cost analytics as described in the skill's purpose.
[SAFE]: Tool coordination descriptions (e.g., using bash for plotting visualizations or grep for log analysis) describe standard operational procedures for the agent to use its existing capabilities for reporting purposes, with no signs of arbitrary or untrusted code execution.

reasoning-controls